Yet another custom V3s board


#1

I have decided to start a new thread for my custom board, since it is a lot earlier in the planning stage.

I wonder how well the driver support are, for this planned custom V3s board:

  • Boots from SPI Flash if there is no bootable microSD card,
  • Wi-Fi over SDIO,
  • Bluetooth over UART (with hardware flow control)
  • 1024x600 7-in capacitive touch screen over parallel display output and I2C for touch panel
  • OV5640 camera on MIPI CSI2
  • Wired Ethernet

And for the overall system, here are also a few questions:

  • Is it worth it to forgo AXP203 and use separate regulators (mostly LP2980 linear and LM3671 switchers) for the board? If so is it a good idea to generate the voltage at the closest possible spot from the chip pins? What supply rail should I use?
  • How to generate the 5V required for the USB OTG operation?
  • Can it be done using a 2-layer board and single-side loading?

#2

I was too thinking about a custom board, and here are my thoughts

  • I would use separate regulators to help guarantee part availability, and sharing among my other designs. Probably switchers for most things, except for the ADC reference.
  • OTG power. Good question. My design would have power through a supply > 5v, so that would be handled.
  • I would just go 4 layer to get the ground and power planes - 4 layer is cheap enough now days. Seems like single side loading is feasible based on @petit_miner 's design.

SDIO wifi - what module choices would be easy to implement aside what was designed in for the Zero?


#3
  • For my design there are only three linear regulators across the entire board: audio analog supply, camera analog supply and backup voltage. Everything else is switch mode.
  • My device is battery operated… I am thinking about just throw in a 5V boost converter to supply most of the system as well as USB (using a MOSFET power switch,) except audio analog and LCD backlight which is direct LDO and boost converter from battery respectively.
  • I am pondering about 4 layer - maybe I will do it if I cannot sort out the power rails.
  • I am using RTL8723BS module for combined SDIO Wi-Fi and UART Bluetooth.

#4

Current distribution is:

  • 1.6A for 1.2V core supply
  • 1.0A for 1.8V DDR supply
  • 0.2A for 30.V AVDD/PLL supply
  • 1.2A for 3.3V supply
  • 0.03A for always-on 3.3V supply

Thus using LDOs, you will get 2W to dissipate if you go from 5V to 3.3V, and 6W from 5V to 1.2V directly, or 3.4W and 3.3W if you cascade them… In any case, you will spend ~6.7W to 8W to heat the birds…

You’d better use DC/DCs or an integrated PMIC, like the AXP203.

If you can’t get an AXP203 (you can get them by ordering V3s on Aliexpress), you can get separate AXP209 that are similar, easy to find individually on Aliexpress:
https://www.aliexpress.com/wholesale?catId=0&initiative_id=SB_20171012235716&SearchText=axp209

For 5V OTG supply, check the Allwinner reference design on page 11:

The V3s is designed to be routed with few layers (2 if you don’t care about EMI or 4 if you care and add inner GND and supply planes), but components on both sides are almost required for all the decoupling capacitors that must be placed close to the numerous supply pins. If you only use single-sided components, you will have a lot of trouble breaking out signals otherwise.

Be careful, it looks like the camera support is not existent at the moment.


#5

Thanks for that.

i couldn’t find any current consumption information in the datasheet; where did this come from?

Interesting on the note about the camera - I’ll post a forum thread separately about that.


#6
  • 1.6A for 1.2V core supply
  • 1.0A for 1.8V DDR supply
  • 0.2A for 30.V AVDD/PLL supply
  • 1.2A for 3.3V supply
  • 0.03A for always-on 3.3V supply

That is a lot of juice… Well this does mean I would have to rework my power delivery.

I just want to skip the AXP-family chips since I have trouble hand soldering QFN and BGA chips. This board will be hand assembled.

My board is powered by a Li-po battery. The battery voltage is boosted to 5V and then bucked down to 3.3V, all using 3A-rated switchers (TPS61030 boost and TPS563201 buck). The 1.2V and 1.8V rails have to come from an LM2642 switcher controller and a few MOSFETs to evenly distribute the power draw from the 3.3V rail. The 3.0V rail is generated using a linear LDO from the 3.3V rail, and the always-on 3.3V rail is generated using a switched-capacitor buck-boost converter REG710NA-33 directly from the battery. This design allows the battery voltage to go all the way down to 2.4V or lower, potentially allow me to discharge the battery deeper and make a better battery life out of it.

The V3s is designed to be routed with few layers (2 if you don’t care about EMI or 4 if you care and add inner GND and supply planes), but components on both sides are almost required for all the decoupling capacitors that must be placed close to the numerous supply pins. If you only use single-sided components, you will have a lot of trouble breaking out signals otherwise.

I will have to make single-side loading work as the PCB is also used to hold the frame for the main display, Caps on the back will short out, and chips will have thermal issues.


#7

Yes, these are the max values, but it is almost proportional to the operating frequency.

QFN chips are actually easy to solder using an hot air station, and you can find good ones <$40 on Aliexpress:
https://www.aliexpress.com/item/sku/32817154703.html

Here are my recommendations, take them into consideration if you want, I am just trying to help and avoid you some very bad surprises.

Going down to 2.4V on a Lipo battery is really a bad idea, you will likely kill it very quickly, see:

https://learn.sparkfun.com/tutorials/battery-technologies

https://www.rcgroups.com/forums/showthread.php?2015863-LiPo-How-low-can-you-go-(voltage-per-cell)

A Lipo battery is considered empty at 3.0V, going below will destroy it, and the slope from 3.3V down to 3.0V is very steep, so a boost converter won’t be very useful in this (very short) area.

Instead of using a costly buck/boost DC/DC, you should rather try to get the best possible efficiency by using good power inductors and DC/DC chips, and TI ones are far from being the cheapest, consider MPS or Diodes Inc. brands instead.

Separate chips will always cost more and take mroe real estate than an integrated PMIC like the AXP203, and these chips have the advantage of featuring a nice power gauge (Coulomb counter) that will provide an accurate (not only voltage-based) remaining capacity indication, allowing you to get the most out of the battery much like a cell phone.

As for single-sided loading, it is also a very bad idea, since the V3s is MEANT to be used with double-sided components. If not, the 30+ required decoupling/bulk capacitors will be too far from power pins to be useful (considering the high power current draws) and will likely cause the V3s to work unreliably because of voltage droops under load…

You can always use spacers and/or isolation sheet to avoid short-circuits, and you can reduce PCB thickness down to 1.2 mm / 1.0 mm to compensate for the additional component height on the second side.


#8

For my double-layer board the pin layout of AXP203 isn’t exactly convenient for me. I fully understand that TI isn’t the cheapest brand for regulators but theirs are more reliable and can take more of a beating than others’.

As of the battery cut-off voltage, without the boost converter AXP203 often cuts off at 3.5V or so wasting a significant portion of the battery’s capacity. My board also have an always-on microcontroller which is used to keep the cut-off voltage (and power sequence, and something more) correct.


#9

You are speaking about convenience, I am talking about reliability… Better have a less convenient board that is working than a convenient board in a drawer :wink:
Anyway, I don’t see why the AXP203 is not convenient, see this (singled-sided loaded) DRC/ERC checked layout:

Open your mind, TI golden days are now long gone, and they are now too fat to innovate and need more cash to sustain their carcass, so they are now increasing their prices and bypassing distributors to get more margin… As for innovation and reliability, MPS uses a patented flip-chip technology that provides a much better power dissipation, resulting in less heat for the same power, as well as QFN chips with integrated inductors, which provide the tightest current loop, considerably lowering EMIs (I am in no way related to them). As for TI legendary support, it is also a thing from the past, see their CC3300 SmartConfig fiasco in direct in this thread: http://e2e.ti.com/support/wireless_connectivity/simplelink_wifi_cc3000/f/851/t/253463

Diodes Inc. also feature some very interesting chips: lines are moving, don’t stick to old prejudices! Do not put affective bias and/or traditions into your technical choices, reality and cold effectiveness will always bite you otherwise.

As for the AXP203 cu-off voltage, it is completely adjustable, as well as power sequencing, so that you don’t need an additional always-on MCU. And looking at the second curve above for 1C (in light blue), 3.5V is ~92% of the battery capacity, so you are not “wasting a significant portion” of its capacity, just ~8% at 1C, more like 3% at 0.5C anyway.

Again, you asked for advice, but in the end it is your design, I am just trying to avoid complete disaster.


#10

I’m also looking for a DC to DC buck converter, I found the AP3410.
Fortunately board size isn’t an issue for my project. I will use 3 AP3410 for the rails 3,3V 1,8V and 1,2V.
I will use a LDO for the 3V rail.
Do you think this is a good decision?


#11

My plan was using a whole bunch of TPS563201’s. I am reinvestigating the AXP203 as well as investigating using a microcontroller-based SMPS (STM32F334 + IR2101 + MOSFETs)


#12

Thank you, the TPS… looks very interesting and is easier to buy than the AP3410 or the AXP PMICs.
Does anyone know how to get the V3s beside buying it at Aliexpress?


#13

I don’t think there is any other way to buy V3s. Allwinner does not bother responding to DIYers, and they don’t make their chips available through major international distributors like Mouser or Farnell.


#14

That isn’t nice, but we can’t do anything against it.
I bought all of my Allwinner V3s in a neighborhood shop just around the corner :smile:
Olimex sells a lot of Allwinner ICs, but not the V3s. The even sell APX PMICs.


#15

I noticed that you mentioned having the datasheet for the axp203 PMC.
I’ve been unable to find a datasheet for it.
Anyone have a tip on where to find it?


#16

AXP203 is mostly the same as AXP209 , could be some differences in default voltages etc.
AXP203 datasheet you probably could find on github SushiBits/SushiBitsCamera …


#17

AXP203 datasheet is here:

AXP209 datasheet is here:
http://dl.linux-sunxi.org/AXP/AXP209_Datasheet_v1.0en.pdf


#18

That SushiBits account is mine… I had to pester Allwinner and X-Power for quite a but to get them fess up the schematics. It was two of my most painful telephone calls ever.


#19

Sorry, I did not know that, I guess it will bring you some traffic :wink: